Empirical Analysis of GNN Architectures for VM Power Consumption Prediction

Authors

Morris Kaburu, Dedan Kimathi University of Technology, Kenya

Patrick Gikunda, Dedan Kimathi University of Technology, Kenya

Juliet Moso, Dedan Kimathi University of Technology, Kenya

Published in

IST-Africa 2026 Conference Proceedings

ISSN: 2576-8581

ISBN: 978-1-905824-76-2

DOI: https://doi.org/10.67725/IST-Africa.2026.OFPX9691

Publisher

IST-Africa Institute and IIMC International Information Management Corporation Ltd

Published in Ireland

Abstract

Power consumption prediction for virtual machines (VMs) is key in improving energy efficiency, capacity planning and sustainability in cloud computing. In this work, we present a comprehensive empirical study that compares several Graph Neural Network (GNN) architectures such as Graph Convolutional Network (GCN), GraphSAGE, Graph Attention Network (GAT), a Temporal GAT / dynamic graph design, GIN, R-GCN, Graph Transformer and Graph NODE for predicting VM power consumption using a real VM monitoring dataset. We assess the impact of different graph construction strategies including k-nearest neighbour, feature similarity threshold and temporal adjacency on predictive performance. Using standard regression metrics (MAE, MSE, RMSE, R²), we reveal that while all GNN models struggle to consistently surpass a naïve mean baseline, feature similarity threshold provides better performance and T-GAT architecture perform better than other. We analyze failure modes and offer recommendations for future work, including richer feature engineering, learned graph structures, and hybrid temporal-spatial modelling.

Keywords

Graph Neural Networks, VM Power Prediction, GCN, GraphSAGE, GAT, Temporal GNN, GIN, T-GATs, R-GCN, Graph Node, Cloud Energy, Spatio-Temporal Forecasting

Cite this paper

M. Kaburu, P. Gikunda and J. Moso (2026) "Empirical Analysis of GNN Architectures for VM Power Consumption Prediction", IST-Africa 2026 Conference Proceedings, Miriam Cunningham and Paul Cunningham (Eds), IST-Africa Institute and IIMC, 2026, ISSN: 2576-8581, ISBN: 978-1-905824-76-2, https://doi.org/10.67725/IST-Africa.2026.OFPX9691

Access Paper

Click here to access this paper

IST-Africa Community Members can download papers by logging into their existing account.

If you wish to become a member of IST-Africa Community, please register your profile and you will then receive login details which you can use to access the papers.

Access to papers in the IST-Africa 2026 Conference Proceedings is for INDIVIDUAL USE only. Neither the Conference Proceedings (or individual double blind peer reviewed papers) are to be included in any public or institutional repository or processed in the context of any AI tool or large language model without prior written permission from the Publishers.

While the IST-Africa Institute and IIMC retain copyright of the Conference Proceedings, each author retains copyright to their individual paper. No paper or part thereof may be reproduced without the written permission of both the Conference Proceedings Publishers and appropriate author(s).